Dynamic prediction of risk levels for manufacturing operations through leading risk indicators: autonomous multivariate risk analyzer method and system

Abstract

The invention provides a dynamic risk analyzer (DRA) that periodically assesses real-time or historic process data, or both, associated with a plant operation, such as a manufacturing, production or processing facility, and identifies hidden near-misses of such operation, when in real time the process data appears otherwise normal. DRA assesses process data to enable operating personnel including management at a facility, to have a comprehensive understanding of risk status and changes in alarm and non-alarm variables. Hidden process near-miss data may be analyzed alone or in combination with other process data and/or data resulting from prior near-miss situations, to identify strategic action to be taken to reduce the severity of, or avert, adverse incidents or catastrophic failure of a facility operation. The invention also provides autonomous learning and monitoring of historical inter variable relationships to distinguish normal multivariate behavior from anomalous departures indicative of undesirable process issues or sensor asynchronicity.

We claim: 
     
         1 . A dynamic system for analyzing risk levels for a manufacturing operation by a user, the system comprising:
 a server that receives
 at least two operation variables comprising automatically measured process data from a real-time data source, 
 a historical-archive data source of the variables or a long-term process data source of the variables defining a period preceding the automatically measured process data, and 
 a historical relationship autonomously learned among the time series data comprising a collection of historical variables, previously uploaded to the server by the user, or an agent or employee of the user; 
   a processor that
 compares the operation variables to the historical-archive data and historical relationship, 
 identifies operation variables that are consistent with normal inter-variable behavior from anomalous departures in multivariate relationships within a range of tolerance, 
 updates the identification of the operation variables over time to identify operational risk and/or near-miss risk that would otherwise be unknown or concealed in the operation variables; and 
   a display that presents the operational risk and/or near-miss risk in a graphic that visually depicts the operational risk and/or near-miss risk value for the variables over a previous time horizon designated by the user or an agent or employee of the user;   wherein the system continuously and autonomously operates contemporaneously with the manufacturing operation.   
     
     
         2 . The system of  claim 1 , wherein the historical relationship is non-linear. 
     
     
         3 . The system of  claim 1 , wherein the process data is either an analog signal or a two-state signal. 
     
     
         4 . The system of  claim 1 , wherein the process data is not communicated outside the manufacturing operation. 
     
     
         5 . The system of  claim 1 , wherein the system operates perpetually without an operator after the historical relationship has been received. 
     
     
         6 . A method for dynamic prediction of risk levels in a manufacturing operation comprising:
 collecting: (1) at least two operation variables comprising automatically measured process data from a source collected by a data-collection component, located within the manufacturing operation, in either (a) real-time or (b) from an archive server or both, (2) long-term process data for a period preceding the collecting measured data, and (3) a historical relationship autonomously learned among the long-term process data, previously uploaded to the server by the user, or an agent or employee of the user;   identifying risk and/or near-miss risk of the manufacturing operation that would otherwise be unknown or concealed by (1) comparing the operation variables to the historical-archive data and historical relationship, (2) identifying operation variables that are consistent with normal inter-variable behavior from anomalous departures in multivariate relationships within a range of tolerance, and (3) updating the identification of the operation variables over time to identify operational risk and/or near-miss risk; and   displaying the risk or near-miss risk in a graphic that visually depicts the operational risk and/or near-miss risk value for the variables over a previous time horizon designated by the user or an agent or employee of the user;   wherein the method is performed continuously and autonomously.   
     
     
         7 . The method of  claim 6 , wherein the historical relationship is non-linear. 
     
     
         8 . The method of  claim 6 , wherein the process data is either an analog signal or a two-state signal. 
     
     
         9 . The method of  claim 6 , wherein the process data is not communicated outside the manufacturing operation. 
     
     
         10 . The method of  claim 6 , wherein the method operates perpetually without an operator after the historical relationship has been collected. 
     
     
         11 . A display system for risk indicators for a manufacturing operation comprising:
 identifying risk and/or near-miss risk of the manufacturing operation that would otherwise be unknown or concealed by (1) comparing the operation variables comprising automatically measured process data from a real-time data source, to the historical-archive data and historical relationship, (2) identifying operation variables that are consistent with normal inter-variable behavior from anomalous departures in multivariate relationships within a range of tolerance, and (3) updating the identification of the operation variables over time to identify operational risk and/or near-miss risk;   using a historical relationship autonomously learned among the time series data comprising a collection of historical variables, previously uploaded to the server by the user, or an agent or employee of the user; and   displaying the risk or near-miss risk in a graphic that visually depicts the operational risk and/or near-miss risk value for the variables over a previous time horizon designated by the user or an agent or employee of the user.   
     
     
         12 . The system of  claim 11 , wherein the historical relationship is non-linear. 
     
     
         13 . The system of  claim 11 , wherein the process data is either an analog signal or a two-state signal. 
     
     
         14 . The system of  claim 11 , wherein the process data is not communicated outside the manufacturing operation. 
     
     
         15 . The system of  claim 11 , wherein the system operates autonomously. 
     
     
         16 . The system of  claim 11 , wherein the system operates perpetually without an operator after the historical relationship has been compared.